Loss Aversion: The 2.25 Coefficient, and How Often a Stock Holder Sees a Loss
Loss aversion is the tendency for a loss to count for more in a decision than a gain of the same size. In Tversky and Kahneman's 1992 experiment the median person weighed losses 2.25 times as heavily as gains, and would accept an even-chance bet of losing $100 only if it could win about $200.
Loss aversion is why losing $500 can sour a week more than winning $500 lifts it. The idea comes from two papers, and the second one attached a number to it. This page goes back to the papers for that number, then measures how often an ordinary index holder is shown a loss, which is where the idea meets a trading account.
How it works
People judge outcomes as gains and losses from where they stand. A trade is not felt as a change in total wealth but as a step up or down from a starting point, usually the entry price or the balance this morning. That starting point is called the reference point.
Below the reference point, the scale gets steeper. Losing a given amount lowers satisfaction by more than winning the same amount raises it. The loss-aversion coefficient is the ratio between the two: at 2.25, losing $100 counts as much as winning $225.
It matters most when a choice mixes the two. Among bets that can only lose, the coefficient scales every option by the same amount and leaves their ranking unchanged. It changes decisions when a bet can go either way, which describes almost every trade.
The two papers behind the number
The 1979 paper introduced the idea. Daniel Kahneman and Amos Tversky’s “Prospect Theory: An Analysis of Decision under Risk,” in Econometrica, volume 47, number 2 (March 1979), proposed that people value gains and losses rather than final wealth. Its abstract states that the value function is generally steeper for losses than for gains, which is loss aversion in one sentence.
The 1992 paper put a figure on it. Tversky and Kahneman’s “Advances in prospect theory: Cumulative representation of uncertainty,” in the Journal of Risk and Uncertainty, volume 5 (October 1992), names loss aversion as one of its two guiding principles. Fitting a model to each of 25 subjects’ choices, the authors report on page 311 a median exponent of 0.88 for the value of both gains and losses and a median loss-aversion coefficient of 2.25. The 25 were graduate students at Berkeley and Stanford, paid a flat $25 rather than according to their choices.
Their Table 6, on page 312, shows it without a model. Subjects named the gain that would make an even chance of losing a fixed amount as attractive as nothing at all. The median answers were $61 to offset a possible $25 loss, $101 for $50, $202 for $100 and $280 for $150, ratios of 2.44, 2.02, 2.02 and 1.87. The authors read this as showing that an even-chance bet becomes acceptable only when the gain is at least twice the loss.
Why checking often makes it worse
Loss aversion is felt each time a result is looked at. Shlomo Benartzi and Richard Thaler combined it with the habit of evaluating a portfolio frequently and called the pair myopic loss aversion, in NBER Working Paper 4369 (May 1993), later published in the Quarterly Journal of Economics (1995). Their abstract reports that the extra return stocks have paid over bonds fits prospect theory’s estimated parameters if investors evaluate their portfolios about once a year.
The market’s own record shows why the checking interval matters. The same holding produces losses at very different rates depending on how often it is examined, which the data below measures.
A worked example
Take an even-chance bet: lose $100, or win an amount X. Someone with no loss aversion would accept any X above $100. With a coefficient of 2.25 and the value of money otherwise treated as a straight line, the bet is only worth taking when X is at least $100 x 2.25 = $225.
The paper’s own exponent raises it a little. With the median exponent of 0.88 applied to both sides, the gain has to satisfy X^0.88 = 2.25 x 100^0.88, so X = 100 x 2.25^(1 / 0.88) = $251.31. This leaves out the paper’s separate weighting of probabilities, so treat it as arithmetic on the two published medians, not as the paper’s prediction.
What people actually answered was $202. That is problem 3 in Table 6. The 2.25 was fitted across every problem each subject faced, so it need not match any single question, and the direct answers ran between 1.87 and 2.44 times the loss.
The original data
The data: SPY’s dividend-adjusted daily closes from 29 Jan 1993 to 25 Sep 2026 from Yahoo Finance (downloaded 26 Sep 2026), measured at seven checking intervals. For each one the count is how many periods ended below where they started, dividends included. The counts are in a CSV of losing periods by horizon.
Checked daily, the index showed a loss almost half the time. SPY lost money on 3,831 of 8,471 trading days, 45.2%, and was exactly flat on 62. Over rolling five-day spans the share was 41.0%.
Checked monthly or yearly, far less. It lost in 139 of 403 calendar months, 34.5%, and in 6 of 32 calendar years from 1994 to 2025 (18.8%): 2000, 2001, 2002, 2008, 2018 and 2022. The worst was 2008, at -36.79%.
Over long spans, losses became rare. Rolling 252-day spans lost 18.2% of the time, five-year spans (1,260 trading days) 15.2%, and ten-year spans (2,520 trading days) 8.0%, 476 of 5,952.
Put next to a coefficient of about 2, the checking interval changes how holding feels. A daily checker meets a loss on close to half of all looks, and on those numbers each one weighs about twice an equal gain. A holder who looks once a year meets a loss in fewer than one year in five. The holding is identical; only the number of losses seen differs.
When it fails
Treating 2.25 as a law fails. It is the median of 25 students in one 1992 experiment whose pay did not depend on their choices, and the median direct answers ran from 1.87 to 2.44 times the loss. A coefficient from a lab says little about how any one trader will act.
Blaming every bad exit on it fails. Some losing trades are held because the plan says to hold them. Loss aversion explains a pattern of choices, not a single decision, and a trading psychology problem is often a position-size problem in disguise.
Using it to justify never taking a loss fails. Refusing to realize a loss does not undo it. If losses feel about twice as heavy, the fix is a smaller loss decided in advance: a stop loss placed before entry and a position size that keeps it tolerable.
Reading the SPY counts as a promise fails. They describe one fund from 1993 to 2026. A single stock, a leveraged product or a shorter record can lose over long spans too, and the ten-year figure says nothing about the next ten years.
Related
Trading psychology shows why losing streaks are ordinary and how size decides how they feel, and how to control trading psychology turns that into rules. The stop loss page covers where to put an exit decided in advance, and risk per trade covers how large each loss should be allowed to get. For what a run of losses does to an account, see drawdown.
Decide the exit on a losing trade before you enter, while there is nothing to lose yet. I set the loss I will accept and the size that keeps it small, because once the position is red, the part of me that hates realizing a loss gets a vote it should not have.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.